[ 
https://issues.apache.org/jira/browse/HIVE-3915?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13557789#comment-13557789
 ] 

Hudson commented on HIVE-3915:
------------------------------

Integrated in Hive-trunk-hadoop2 #72 (See 
[https://builds.apache.org/job/Hive-trunk-hadoop2/72/])
    HIVE-3915 Union with map-only query on one side and two MR job query on the 
other
produces wrong results (Kevin Wilfong via namit) (Revision 1435203)

     Result = FAILURE
namit : http://svn.apache.org/viewcvs.cgi/?root=Apache-SVN&view=rev&rev=1435203
Files : 
* /hive/trunk/ql/src/java/org/apache/hadoop/hive/ql/Driver.java
* /hive/trunk/ql/src/java/org/apache/hadoop/hive/ql/optimizer/GenMRUnion1.java
* /hive/trunk/ql/src/test/queries/clientpositive/union33.q
* /hive/trunk/ql/src/test/results/clientpositive/union33.q.out

                
> Union with map-only query on one side and two MR job query on the other 
> produces wrong results
> ----------------------------------------------------------------------------------------------
>
>                 Key: HIVE-3915
>                 URL: https://issues.apache.org/jira/browse/HIVE-3915
>             Project: Hive
>          Issue Type: Bug
>          Components: Query Processor
>    Affects Versions: 0.11.0
>            Reporter: Kevin Wilfong
>            Assignee: Kevin Wilfong
>             Fix For: 0.11.0
>
>         Attachments: HIVE-3915.1.patch.txt
>
>
> When a query contains a union with a map only subquery on one side and a 
> subquery involving two sequential map reduce jobs on the other, it can 
> produce wrong results.  It appears that if the map only queries table scan 
> operator is processed first the task involving a union is made a root task.  
> Then when the other subquery is processed, the second map reduce job gains 
> the task involving the union as a child and it is made a root task.  This 
> means that both the first and second map reduce jobs are root tasks, so the 
> dependency between the two is ignored.  If they are run in parallel (i.e. the 
> cluster has more than one node) no results will be produced for the side of 
> the union with the two map reduce jobs and only the results of the other side 
> of the union will be returned.
> The order TableScan operators are processed is crucial to reproducing this 
> bug, and it is determined by the order values are retrieved from a map, and 
> hence hard to predict, so it doesn't always reproduce.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Reply via email to